platform/upstream/gstreamer.git
7 years agodashdemux: Remove wrong assertion
Edward Hervey [Wed, 31 May 2017 08:58:39 +0000 (10:58 +0200)]
dashdemux: Remove wrong assertion

This is wrong because:
* If the rate is negative we should check for the *previous* period
* adaptivedemux already does the proper checks before calling this
  method

7 years agottml: Simplify code
Edward Hervey [Fri, 26 May 2017 15:55:44 +0000 (17:55 +0200)]
ttml: Simplify code

n2 can never be NULL since:
* it's in a "while (n1 && n2)" block
* and it's not modified before

CID #1405868

7 years agononstreamaudio: Avoid using wrong variable
Edward Hervey [Fri, 26 May 2017 15:44:40 +0000 (17:44 +0200)]
nonstreamaudio: Avoid using wrong variable

And to make that 100% obvious, only use variables declared within the
switch cases instead of function-wide ones.

Also remove useless one-time-use-only variable.

CID #1409857

7 years agononstreamaudio: Avoid using un-initialized value
Edward Hervey [Fri, 26 May 2017 15:41:39 +0000 (17:41 +0200)]
nonstreamaudio: Avoid using un-initialized value

We would end up with cur_position not set but proceed being at its
default value of TRUE. Instead, properly set proceed to FALSE

CID #1409855

7 years agoadaptivedemux: Fix debugging message
Edward Hervey [Fri, 26 May 2017 15:30:10 +0000 (17:30 +0200)]
adaptivedemux: Fix debugging message

GstSegment position is a guint64 and not a gint64

CID #1409910

7 years agoadaptivedemux: Check live seeking range more often
Edward Hervey [Thu, 25 May 2017 07:48:53 +0000 (09:48 +0200)]
adaptivedemux: Check live seeking range more often

The live seeking range was only checked when doing actual seeks. This was
assuming that the rate would always be 1.0 (i.e. the playback would
advance in realtime, and therefore fragments would always be available
since the seeking window moves at the same rate).

With non-1.0 rates, this no longer becomes valid, and therefore we need
to check whether we are still within the live seeking range when advancing.

https://bugzilla.gnome.org/show_bug.cgi?id=783075

7 years agoadaptivedemux: Don't create invalid event
Edward Hervey [Thu, 25 May 2017 14:42:03 +0000 (16:42 +0200)]
adaptivedemux: Don't create invalid event

tags could potentially be NULL

7 years agoadaptivedemux: retry download MAX_DOWNLOAD_RETRY_COUNT times before erroring
Matthew Waters [Fri, 2 Dec 2016 06:51:57 +0000 (17:51 +1100)]
adaptivedemux: retry download MAX_DOWNLOAD_RETRY_COUNT times before erroring

What we want is to retry downloading the fragment on 4xx/5xx errors
however returning EOS will cause waiting for a manifest update for live
(which may be a really long time) or stop everything for non-live.

Change that to only return EOS/ERROR once we've reached the error limit.

https://bugzilla.gnome.org/show_bug.cgi?id=776609

7 years agogl: Fix indentation
Sebastian Dröge [Thu, 25 May 2017 10:06:03 +0000 (13:06 +0300)]
gl: Fix indentation

7 years agoglframebuffer: #define GL_READ_FRAMEBUFFER / GL_DRAW_FRAMEBUFFER if not defined yet
Sebastian Dröge [Thu, 25 May 2017 10:05:23 +0000 (13:05 +0300)]
glframebuffer: #define GL_READ_FRAMEBUFFER / GL_DRAW_FRAMEBUFFER if not defined yet

Just like we do elsewhere already.

7 years agoglformat: fix the usage of GST_GL_RGB565
Haihua Hu [Thu, 25 May 2017 03:05:47 +0000 (11:05 +0800)]
glformat: fix the usage of GST_GL_RGB565

GL_RGB565 is sized internal glformat, the corresponding glformat
should be GL_RGB and type is GL_UNSIGNED_SHORT_565. Otherwise will
return GL_INVALID_ENUM when creating texture.

https://bugzilla.gnome.org/show_bug.cgi?id=783066

7 years agoglframebuffer: check frame buffer status need use specific fbo target
Haihua Hu [Thu, 25 May 2017 02:09:04 +0000 (10:09 +0800)]
glframebuffer: check frame buffer status need use specific fbo target

https://bugzilla.gnome.org/show_bug.cgi?id=783065

7 years agoopenmpt: Fix compilation with 0.2.7386 as in Debian
Sebastian Dröge [Wed, 24 May 2017 08:47:47 +0000 (11:47 +0300)]
openmpt: Fix compilation with 0.2.7386 as in Debian

The OPENMPT_API_VERSION_AT_LEAST macro does not exist.

7 years agotests: Make audiomixer test_clip verify the resulting timestamps too
Olivier Crête [Mon, 22 May 2017 21:06:01 +0000 (23:06 +0200)]
tests: Make audiomixer test_clip verify the resulting timestamps too

7 years agoaggregator: Process serialized queries through the queue
Olivier Crête [Mon, 22 May 2017 22:53:57 +0000 (00:53 +0200)]
aggregator: Process serialized queries through the queue

This ensures that they really get processed in order with
buffers. Just waiting for the queue to be empty is sometimes not
enough as the buffers are dropped from the pad before the result is
pushed to the next element, sometimes resulting in surprising
re-ordering.

7 years agoaggregator: Set flow to FLUSHING on pad stop
Olivier Crête [Mon, 22 May 2017 22:53:23 +0000 (00:53 +0200)]
aggregator: Set flow to FLUSHING on pad stop

Fixes a rare race where the pad is being stopped while doing a query.

7 years agotests: audiointerleave: Remove drain with manual clock
Olivier Crête [Mon, 22 May 2017 22:52:27 +0000 (00:52 +0200)]
tests: audiointerleave: Remove drain with manual clock

Now that the queries go onto the queue, you may need to pull the crank
in order for them to be processed, making this test difficult.

7 years agoRequire wildmidi >= 0.4
Tim-Philipp Müller [Tue, 23 May 2017 08:57:28 +0000 (09:57 +0100)]
Require wildmidi >= 0.4

https://bugzilla.gnome.org/show_bug.cgi?id=768576

7 years agowildmidi: include stdint.h for int8_t
Tim-Philipp Müller [Tue, 23 May 2017 08:17:08 +0000 (09:17 +0100)]
wildmidi: include stdint.h for int8_t

7 years agowin32: update .def file for new base class
Tim-Philipp Müller [Tue, 23 May 2017 07:42:46 +0000 (08:42 +0100)]
win32: update .def file for new base class

https://bugzilla.gnome.org/show_bug.cgi?id=768576

7 years agowildmididec: explicitly cast buffer data to int8 in _decode()
Reynaldo H. Verdejo Pinochet [Mon, 22 May 2017 22:52:33 +0000 (15:52 -0700)]
wildmididec: explicitly cast buffer data to int8 in _decode()

Fixes compiler warning introduced in commit ff32a4297:

gstwildmididec.c:637:47: error: pointer targets in passing argument 2 of ‘WildMidi_GetOutput’ differ in signedness
       WildMidi_GetOutput (wildmidi_dec->song, (char *) (info.data), info.size);
                                               ^
wildmidi_lib.h:106:15: note: expected ‘int8_t * {aka signed char *}’ but argument is of type ‘char *’
 WM_SYMBOL int WildMidi_GetOutput (midi *handle, int8_t *buffer, uint32_t size);

https://bugzilla.gnome.org/show_bug.cgi?id=768576

7 years agoopenmpt: Add openmptdec element
Carlos Rafael Giani [Wed, 27 Jul 2016 00:22:26 +0000 (02:22 +0200)]
openmpt: Add openmptdec element

https://bugzilla.gnome.org/show_bug.cgi?id=768576

7 years agowildmidi: Port to 1.0 on top of the nonstreamaudiodecoder base class
Carlos Rafael Giani [Wed, 8 Mar 2017 21:17:41 +0000 (22:17 +0100)]
wildmidi: Port to 1.0 on top of the nonstreamaudiodecoder base class

https://bugzilla.gnome.org/show_bug.cgi?id=768576

7 years agoaudio: Add nonstreamaudiodecoder base class
Carlos Rafael Giani [Tue, 26 Jul 2016 23:41:20 +0000 (01:41 +0200)]
audio: Add nonstreamaudiodecoder base class

https://bugzilla.gnome.org/show_bug.cgi?id=768576

7 years agogltestsrc: Fix typo (occured->occurred)
Anders Jonsson [Mon, 22 May 2017 11:46:36 +0000 (13:46 +0200)]
gltestsrc: Fix typo (occured->occurred)

https://bugzilla.gnome.org/show_bug.cgi?id=782947

7 years agomeson: don't need config.h.meson any longer
Tim-Philipp Müller [Sun, 21 May 2017 17:03:02 +0000 (18:03 +0100)]
meson: don't need config.h.meson any longer

7 years agoaudioaggregate: Don't hold object locks across calls to aggregate_one
Olivier Crête [Sat, 20 May 2017 17:00:23 +0000 (19:00 +0200)]
audioaggregate: Don't hold object locks across calls to aggregate_one

https://bugzilla.gnome.org/show_bug.cgi?id=782878

7 years agoaggregator: Request pad templates which are not request pad
Olivier Crête [Fri, 18 Nov 2016 19:44:16 +0000 (14:44 -0500)]
aggregator: Request pad templates which are not request pad

https://bugzilla.gnome.org/show_bug.cgi?id=782920

7 years agoaggregator: Don't restrict sink pad names
Olivier Crête [Fri, 18 Nov 2016 19:41:54 +0000 (14:41 -0500)]
aggregator: Don't restrict sink pad names

Sink pads could have other names than sink_%u

https://bugzilla.gnome.org/show_bug.cgi?id=782920

7 years agoglbasemixer: Remove unused negotiated member
Olivier Crête [Sun, 21 May 2017 16:31:59 +0000 (18:31 +0200)]
glbasemixer: Remove unused negotiated member

This is now all handled in GstAggregator, so this code is not
called anymore.

7 years agovideoaggregator: Declare that it supports the video meta on input
Olivier Crête [Sun, 21 May 2017 13:44:02 +0000 (15:44 +0200)]
videoaggregator: Declare that it supports the video meta on input

https://bugzilla.gnome.org/show_bug.cgi?id=782918

7 years agogl*mixer: Use propose_allocation from the GstAggregator base class
Olivier Crête [Sun, 21 May 2017 13:30:10 +0000 (15:30 +0200)]
gl*mixer: Use propose_allocation from the GstAggregator base class

https://bugzilla.gnome.org/show_bug.cgi?id=782918

7 years agoaggregator: Implement propose allocation
Olivier Crête [Sun, 21 May 2017 13:19:17 +0000 (15:19 +0200)]
aggregator: Implement propose allocation

https://bugzilla.gnome.org/show_bug.cgi?id=782918

7 years agoaggregator: Check for the result of caps events
Olivier Crête [Sun, 21 May 2017 12:34:13 +0000 (14:34 +0200)]
aggregator: Check for the result of caps events

https://bugzilla.gnome.org/show_bug.cgi?id=782918

7 years agoaggregator: Caps event always goes to the aggregate thread
Olivier Crête [Sun, 21 May 2017 12:28:00 +0000 (14:28 +0200)]
aggregator: Caps event always goes to the aggregate thread

So no need to check it here.

https://bugzilla.gnome.org/show_bug.cgi?id=782918

7 years agogl/viv-fb: Fix user-choice string comparisons
Carlos Rafael Giani [Tue, 9 May 2017 21:59:04 +0000 (23:59 +0200)]
gl/viv-fb: Fix user-choice string comparisons

https://bugzilla.gnome.org/show_bug.cgi?id=782921

7 years agoqmlglsink: Add dummy texture that is shown as placeholder for NULL buffers
Carlos Rafael Giani [Sun, 21 May 2017 13:26:12 +0000 (15:26 +0200)]
qmlglsink: Add dummy texture that is shown as placeholder for NULL buffers

https://bugzilla.gnome.org/show_bug.cgi?id=782917

7 years agowin32: update .def file for new exports
Tim-Philipp Müller [Sun, 21 May 2017 14:15:48 +0000 (15:15 +0100)]
win32: update .def file for new exports

7 years agovideoaggregator: Create normal video pool as a fallback
Olivier Crête [Sat, 20 May 2017 16:10:29 +0000 (18:10 +0200)]
videoaggregator: Create normal video pool as a fallback

https://bugzilla.gnome.org/show_bug.cgi?id=746529

7 years agoaudioaggregator: Use downstream allocator and params if available
Olivier Crête [Sat, 20 May 2017 15:59:19 +0000 (17:59 +0200)]
audioaggregator: Use downstream allocator and params if available

https://bugzilla.gnome.org/show_bug.cgi?id=746529

7 years agoglbasemixer: Remove own decide_allocation, use GstAggregator's
Olivier Crête [Sat, 20 May 2017 15:35:43 +0000 (17:35 +0200)]
glbasemixer: Remove own decide_allocation, use GstAggregator's

https://bugzilla.gnome.org/show_bug.cgi?id=746529

7 years agoglbasemixer: Use aggregator for allocation handling
Olivier Crête [Sat, 20 May 2017 15:30:06 +0000 (17:30 +0200)]
glbasemixer: Use aggregator for allocation handling

https://bugzilla.gnome.org/show_bug.cgi?id=746529

7 years agovideoaggregator: Get the buffer from the pool if available
Olivier Crête [Sat, 20 May 2017 15:25:16 +0000 (17:25 +0200)]
videoaggregator: Get the buffer from the pool if available

https://bugzilla.gnome.org/show_bug.cgi?id=746529

7 years agoaggregator: Add downstream allocation query
Olivier Crête [Sat, 20 May 2017 14:58:54 +0000 (16:58 +0200)]
aggregator: Add downstream allocation query

https://bugzilla.gnome.org/show_bug.cgi?id=746529

7 years agovideoaggregator: Return to parent on reconfigure
Olivier Crête [Sun, 21 May 2017 10:41:53 +0000 (12:41 +0200)]
videoaggregator: Return to parent on reconfigure

The caps negotiation is now in the parent, so need to return there
if a reconfiguration is needed, otherwise it will loops forever.

7 years agogl: x11: fix compiler warning
Tim-Philipp Müller [Sun, 21 May 2017 10:44:37 +0000 (11:44 +0100)]
gl: x11: fix compiler warning

7 years agoglcontext: add public swap_buffers function
Matthew Waters [Sun, 21 May 2017 10:03:01 +0000 (12:03 +0200)]
glcontext: add public swap_buffers function

That simply calls the implementation

7 years agoexamples: gl: remove ancient bitrotten .vcproj files
Tim-Philipp Müller [Sun, 21 May 2017 09:57:18 +0000 (10:57 +0100)]
examples: gl: remove ancient bitrotten .vcproj files

7 years agoadaptivedemux: fix debug message printf format
Graham Leggett [Sat, 20 May 2017 16:22:10 +0000 (16:22 +0000)]
adaptivedemux: fix debug message printf format

Match gsize to G_GSIZE_FORMAT in the debug message.

https://bugzilla.gnome.org/show_bug.cgi?id=782873

7 years agowin32: update .def file for API removal
Tim-Philipp Müller [Sat, 20 May 2017 16:04:52 +0000 (17:04 +0100)]
win32: update .def file for API removal

7 years agoaudiointerleave: Take object lock while modifying channel count
Olivier Crête [Sat, 20 May 2017 15:47:04 +0000 (17:47 +0200)]
audiointerleave: Take object lock while modifying channel count

7 years agoaggregator: Remove unused GST_FLOW_NOT_HANDLED
Olivier Crête [Sat, 20 May 2017 13:56:16 +0000 (15:56 +0200)]
aggregator: Remove unused GST_FLOW_NOT_HANDLED

7 years agoaggregator: add simple support for caps handling
Matthew Waters [Sat, 20 May 2017 12:24:57 +0000 (14:24 +0200)]
aggregator: add simple support for caps handling

Modelled off the videoaggregator caps handling as that seems the most
mature aggregtor-using implementation that has caps handling there is.

https://bugzilla.gnome.org/show_bug.cgi?id=776931

7 years agoaggregator: Reset upstream latency on first buffer
Nicolas Dufresne [Sat, 20 May 2017 11:10:53 +0000 (13:10 +0200)]
aggregator: Reset upstream latency on first buffer

In the case an aggregator is created and pads are requested but only
linked later, we end up never updating the upstream latency.
This was because latency queries on pads that are not linked succeed,
so we never did a new query once a live source has been linked, so the
thread was never started.

https://bugzilla.gnome.org/show_bug.cgi?id=757548

7 years agovideoaggregator: delay using new caps from a sink pad until the next buffer in the...
George Kiagiadakis [Tue, 4 Apr 2017 08:25:43 +0000 (11:25 +0300)]
videoaggregator: delay using new caps from a sink pad until the next buffer in the queue is taken

When caps changes while streaming, the new caps was getting processed
immediately in videoaggregator, but the next buffer in the queue that
corresponds to this new caps was not necessarily being used immediately,
which resulted sometimes in using an old buffer with new caps. Of course
there used to be a separate buffer_vinfo for mapping the buffer with its
own caps, but in compositor the GstVideoConverter was still using wrong
info and resulted in invalid reads and corrupt output.

This approach here is more safe. We delay using the new caps
until we actually select the next buffer in the queue for use.
This way we also eliminate the need for buffer_vinfo, since the
pad->info is always in sync with the format of the selected buffer.

https://bugzilla.gnome.org/show_bug.cgi?id=780682

7 years agoaggregator: Always handle sync'ed events on output thread
Olivier Crête [Sat, 14 May 2016 13:52:37 +0000 (15:52 +0200)]
aggregator: Always handle sync'ed events on output thread

Having all synchronized events always be handled on the output
thread should make synchronization easier.

https://bugzilla.gnome.org/show_bug.cgi?id=781673

7 years agoaggregator: Delay clipping to output thread
Olivier Crête [Wed, 6 Jul 2016 20:39:17 +0000 (16:39 -0400)]
aggregator: Delay clipping to output thread

This is required because the synchronized events like caps or segments
may only be processed on the output thread.

https://bugzilla.gnome.org/show_bug.cgi?id=781673

7 years agoaggregator: Make pad eos as soon as all buffers are processed, dont way for events
Olivier Crête [Thu, 7 Jul 2016 20:13:57 +0000 (16:13 -0400)]
aggregator: Make pad eos as soon as all buffers are processed, dont way for events

https://bugzilla.gnome.org/show_bug.cgi?id=781673

7 years agoaggregator: Only count buffers when declaring queue full
Olivier Crête [Thu, 7 Jul 2016 15:47:40 +0000 (11:47 -0400)]
aggregator: Only count buffers when declaring queue full

https://bugzilla.gnome.org/show_bug.cgi?id=781673

7 years agotests: Test caps using query
Olivier Crête [Wed, 6 Jul 2016 21:28:11 +0000 (17:28 -0400)]
tests: Test caps using query

Sending an event can accepted event if the caps were rejected
because the event could be queued and processed later.

Also send a drain query in the caps test to make sure that the
event has been processed.

https://bugzilla.gnome.org/show_bug.cgi?id=781673

7 years agoaggregator: Simplify clip function
Olivier Crête [Wed, 6 Jul 2016 20:41:44 +0000 (16:41 -0400)]
aggregator: Simplify clip function

The return value was ignored anyway

https://bugzilla.gnome.org/show_bug.cgi?id=781673

7 years agoaggregator: Only declare first buffer on actual buffer
Olivier Crête [Sun, 15 May 2016 13:04:58 +0000 (16:04 +0300)]
aggregator: Only declare first buffer on actual buffer

The function needs to be unlocked if any data is received, but only
end the first buffer processing on an actual buffer, synchronized events
don't matter on the first buffer processing.

https://bugzilla.gnome.org/show_bug.cgi?id=781673

7 years agoaggregator: Set initial position on first buffer
Olivier Crête [Wed, 10 May 2017 00:20:07 +0000 (20:20 -0400)]
aggregator: Set initial position on first buffer

Set the initial position on the first buffer, otherwise the queue
will grow without limits before the output thread is started.

https://bugzilla.gnome.org/show_bug.cgi?id=781673

7 years agoaggregator: Reset the pad's first buffer flag with the rest
Olivier Crête [Wed, 10 May 2017 00:06:29 +0000 (20:06 -0400)]
aggregator: Reset the pad's first buffer flag with the rest

There is not reason to have separate code to reset this one.

https://bugzilla.gnome.org/show_bug.cgi?id=781673

7 years agoaggregator: Reset pad on init
Olivier Crête [Wed, 10 May 2017 00:05:55 +0000 (20:05 -0400)]
aggregator: Reset pad on init

Factor out the pad reset code from the flushing and use it on init as well

https://bugzilla.gnome.org/show_bug.cgi?id=781673

7 years agoaggregator: Fix indentation
Olivier Crête [Wed, 10 May 2017 00:13:58 +0000 (20:13 -0400)]
aggregator: Fix indentation

https://bugzilla.gnome.org/show_bug.cgi?id=781673

7 years agoglcontext: keep a ref to the active thread
fvanzile [Thu, 23 Feb 2017 23:42:08 +0000 (15:42 -0800)]
glcontext: keep a ref to the active thread

With the macOS/iOS implementations, the active thread can change
multiple times over the life of a pipeline which would expose a race in
the thread tracking.

Fix by taking a ref on the active thread while the context is active.

https://bugzilla.gnome.org/show_bug.cgi?id=779202

7 years agomeson: add options to set package name and origin
Tim-Philipp Müller [Sat, 20 May 2017 14:04:45 +0000 (15:04 +0100)]
meson: add options to set package name and origin

https://bugzilla.gnome.org/show_bug.cgi?id=782172

7 years agoexamples: fix macOS 9.12 deprecation warnings
Josep Torra [Sat, 20 May 2017 10:34:27 +0000 (12:34 +0200)]
examples: fix macOS 9.12 deprecation warnings

Add #defines to allow older versions of macOS to use the new constant names.

7 years agococoa: fix a recently introduced typo
Josep Torra [Sat, 20 May 2017 10:19:08 +0000 (12:19 +0200)]
cocoa: fix a recently introduced typo

Fixes gstgldisplay_cocoa.m:175:26: error: use of undeclared identifier 'singletone'.

7 years agococoa: fix macOS 10.12 deprecation warnings
Josep Torra [Sat, 20 May 2017 10:16:50 +0000 (12:16 +0200)]
cocoa: fix macOS 10.12 deprecation warnings

Add #defines to allow older versions of macOS to use the new constant names.

7 years agodashdemux: Use correct (sub) fragment start when clipping chunk size
Sebastian Dröge [Mon, 17 Apr 2017 11:43:49 +0000 (14:43 +0300)]
dashdemux: Use correct (sub) fragment start when clipping chunk size

If we didn't download anything yet, we shouldn't use fragment.start but
the start position of the current sidx entry.

7 years agodashdemux: Make sure to pass the next buffer after a seek with DISCONT flag
Sebastian Dröge [Mon, 17 Apr 2017 11:25:10 +0000 (14:25 +0300)]
dashdemux: Make sure to pass the next buffer after a seek with DISCONT flag

When we manually seek for skipping ahead in keyunit-only trickmode, we
have to enforce that ourselves as adaptivedemux does not know about the
seeks.

7 years agodashdemux: Ensure that we never ever download over the current SIDX entry
Sebastian Dröge [Sat, 15 Apr 2017 08:55:34 +0000 (11:55 +0300)]
dashdemux: Ensure that we never ever download over the current SIDX entry

Even if downloading more than needed to try to get the moof and first
keyframe both together.

7 years agodashdemux: Take SIDX entry timestamp/duration into account for keyunit-only mode
Sebastian Dröge [Thu, 13 Apr 2017 19:49:29 +0000 (22:49 +0300)]
dashdemux: Take SIDX entry timestamp/duration into account for keyunit-only mode

And not the whole segment's timestamp/duration

7 years agodashdemux: Reset parsing state correctly in various places
Sebastian Dröge [Thu, 13 Apr 2017 14:12:26 +0000 (17:12 +0300)]
dashdemux: Reset parsing state correctly in various places

7 years agodashdemux: Reset previous target_time when seeking
Sebastian Dröge [Thu, 13 Apr 2017 11:51:07 +0000 (14:51 +0300)]
dashdemux: Reset previous target_time when seeking

7 years agodashdemux: Smooth out skip distances in keyframe-only trick modes
Sebastian Dröge [Mon, 3 Apr 2017 13:47:58 +0000 (16:47 +0300)]
dashdemux: Smooth out skip distances in keyframe-only trick modes

This ensures smoother playback. It looks weird if we first do a big
jump, then play a couple of consecutive frames, just to again skip ahead
quite a bit because we ran late again.

7 years agodashdemux: Take configure max bitrate/framerate into account for keyframe skipping
Sebastian Dröge [Mon, 3 Apr 2017 12:51:43 +0000 (15:51 +0300)]
dashdemux: Take configure max bitrate/framerate into account for keyframe skipping

And by default only produce up to 10 fps.

7 years agodashdemux: Try harder to not run into a loop over the same fragment over and over
Sebastian Dröge [Mon, 3 Apr 2017 12:48:13 +0000 (15:48 +0300)]
dashdemux: Try harder to not run into a loop over the same fragment over and over

While still making sure to not jump ahead one fragment further than
needed.

7 years agodashdemux: Add some debug output to the target time selection
Sebastian Dröge [Wed, 15 Mar 2017 15:47:53 +0000 (17:47 +0200)]
dashdemux: Add some debug output to the target time selection

7 years agodashdemux: Use the current clock running time in addition to the QoS earliest time
Sebastian Dröge [Tue, 28 Mar 2017 11:12:12 +0000 (14:12 +0300)]
dashdemux: Use the current clock running time in addition to the QoS earliest time

7 years agodashdemux: Download keyframes from the current position if we're far enough ahead
Sebastian Dröge [Mon, 27 Mar 2017 14:52:36 +0000 (17:52 +0300)]
dashdemux: Download keyframes from the current position if we're far enough ahead

Far enough here means more than 500ms or 4 times the average keyframe
download time. There is no need to jump ahead by one average keyframe
download time in this case.

This makes playback smooth if the network is fast enough.

7 years agodashdemux: Fix sync sample selection based on target time in reverse playback mode
Sebastian Dröge [Wed, 22 Mar 2017 10:05:20 +0000 (12:05 +0200)]
dashdemux: Fix sync sample selection based on target time in reverse playback mode

7 years agodashdemux: Snap-seek for skipping ahead and use actual keyframe distance in the curre...
Sebastian Dröge [Wed, 22 Mar 2017 09:21:47 +0000 (11:21 +0200)]
dashdemux: Snap-seek for skipping ahead and use actual keyframe distance in the current fragment if applicable

7 years agodashdemux: Don't adjust for fragment duration twice when seeking ahead in KEYUNITS...
Sebastian Dröge [Wed, 15 Mar 2017 14:47:59 +0000 (16:47 +0200)]
dashdemux: Don't adjust for fragment duration twice when seeking ahead in KEYUNITS mode

7 years agodashdemux: Add current fragment duration in reverse playback mode to the position
Sebastian Dröge [Wed, 15 Mar 2017 14:46:55 +0000 (16:46 +0200)]
dashdemux: Add current fragment duration in reverse playback mode to the position

We play from the end of the fragment to the beginning, not from the
beginning backwards.

7 years agodashdemux: Never take more than the current fragment duration for estimations inside...
Sebastian Dröge [Wed, 15 Mar 2017 14:45:33 +0000 (16:45 +0200)]
dashdemux: Never take more than the current fragment duration for estimations inside the fragment

Taking the average might give us from results.

7 years agodashdemux: Don't increase current position if we just downloaded the moof in KEYUNIT...
Sebastian Dröge [Mon, 6 Mar 2017 12:27:20 +0000 (14:27 +0200)]
dashdemux: Don't increase current position if we just downloaded the moof in KEYUNIT mode

7 years agodashdemux: Improve key-unit trick mode downloading
Edward Hervey [Thu, 12 Jan 2017 14:54:37 +0000 (15:54 +0100)]
dashdemux: Improve key-unit trick mode downloading

When dealing with key-unit trick mode downloads, the goal is to
provide the best "Quality of Experience". This is achieved by:
1) maximizing the number of frames displayed per second
2) avoiding "stalling" as much as possible (i.e. not downloading and
  decoding frames fast enough)

This implementation achives this by:
1) Knowing very precisely the current keyframe being download (i.e
  more accurate than at the fragment level which might contain more
  than one keyfram). This is the new "actual_position" variable
  introduced by this commit
2) Knowing the position of downstream (provided by QoS and stored
  in the adaptivedemuxstream qos_earliest_time variable)
3) Knowing how long it takes to request and fully download a keyframe
  (the average_download_time variable)

Taking those 3 variables into account, whenever a keyframe has been
pushed downstream we calculate a "target time" (target_time variable)
which is the ideal next keyframe time to request so that:
1) It will be requested/downloaded/demuxed/decoded in time to be
  displayed without being too late
2) It will not be too far ahead that it would cause too few frames
  per second to be displayed.

How far ahead we will request is inversily proportional to how close
the actual position (actual_position) is from the downstream
position (qos_earliest_time). The more is buffered between the source
and the sink, the "closer" the target time will be, and therefore
the more frames per seconds will be displayed (up to the limit
of keyframes_per_second * absolute_rate).

7 years agodashdemux: Store average download time
Edward Hervey [Wed, 11 Jan 2017 16:11:27 +0000 (17:11 +0100)]
dashdemux: Store average download time

This will be used to bound the download rate when working in
keyframe-only trick mode

7 years agoadaptivedemux: Add various comments to the code
Edward Hervey [Wed, 11 Jan 2017 16:08:36 +0000 (17:08 +0100)]
adaptivedemux: Add various comments to the code

7 years agoadaptivedemux: Store QoS time
Edward Hervey [Tue, 15 Nov 2016 07:13:27 +0000 (08:13 +0100)]
adaptivedemux: Store QoS time

Allows subclasses to know where downstream is and make decisions
based upon that

7 years agodashdemux: Collect keyunit distance and size statistics
Sebastian Dröge [Thu, 1 Sep 2016 14:47:11 +0000 (17:47 +0300)]
dashdemux: Collect keyunit distance and size statistics

7 years agodashdemux: Disable bitrate switching in key-unit trick mode
Edward Hervey [Mon, 7 Nov 2016 11:22:09 +0000 (12:22 +0100)]
dashdemux: Disable bitrate switching in key-unit trick mode

This creates too much havoc for now

7 years agodashdemux: Use new adaptivedemux trickmode macro
Edward Hervey [Mon, 31 Oct 2016 09:08:35 +0000 (10:08 +0100)]
dashdemux: Use new adaptivedemux trickmode macro

Reduces the lines of code, and makes it a bit more readable

7 years agoadaptivedemux: Add a macro to know if we are in trickmode-keyunit
Edward Hervey [Mon, 31 Oct 2016 09:08:04 +0000 (10:08 +0100)]
adaptivedemux: Add a macro to know if we are in trickmode-keyunit

Reduces the amount of lines of code in subclasses

7 years agouvch264src: Apply timestamps to outgoing aux buffers
Jan Schmidt [Thu, 18 May 2017 13:57:22 +0000 (15:57 +0200)]
uvch264src: Apply timestamps to outgoing aux buffers

When extracting an aux buffer from an MJPG carrier, at
*least* put the original timestamp on it, even if we
fail to apply any other timestamp (which we always do
at the moment, because the timestamp calculating code
was never finished). Apply a DTS using the camera
supplied delay value as well, assuming that there's
no re-ordering going on (there isn't in the C920,
which is really the only extant camera doing this
stuff) and a warning if that turns out not to be true.

7 years agoqt: Use GST_GL_HAVE_PLATFORM_CGL instead of GST_GL_HAVE_PLATFORM_COCOA
Simon Himmelbauer [Thu, 18 May 2017 12:23:14 +0000 (15:23 +0300)]
qt: Use GST_GL_HAVE_PLATFORM_CGL instead of GST_GL_HAVE_PLATFORM_COCOA

The latter is not used/available anymore since years. Also fix a typo
in the include path for the Cocoa GL display header.

7 years agomotioncells: delay motionmaskcoords until caps arrive
Dmitry Zhadinets [Sun, 7 May 2017 19:50:00 +0000 (19:50 +0000)]
motioncells: delay motionmaskcoords until caps arrive

motionmaskcoords is not applied on start because the information
about resolution isn't available until caps arrive.

https://bugzilla.gnome.org/show_bug.cgi?id=768666